Automatic Maintenance of a Computing System in a Steady State Using Correlation

1. A method for automatically maintaining an autonomic computing system in a steady state, the autonomic computing system having a plurality of parameters, each parameter having one or more thresholds, the autonomic computing system having a plurality of influencers, adjustment of the influencers affecting values of the parameters, the method comprising: in response to determining that one or more of the parameters are each reaching one of the thresholds of the parameter, the one or more of the parameters referred to as to-be-affected parameters, identifying each to-be-affected parameter and the thresholds of the to-be-affected parameter; for each influencer, determining a correlation value between the influencer and each to-be-affected parameter; and, adjusting one or more of the influencers so that the to-be-affected parameters return to more-normal values, based on the correlation values determined, wherein the influencers that are each distinctly associated with the parameters are referred to as distinct influencers, such that adjustment of each distinct influencer affects the value of only one of the parameters, wherein the influencers that are each commonly associated with more than one of the parameters are referred to as common influencers, such that adjustment of each common influencer affects the values of more than one of the parameters, and wherein adjusting one or more of the influencers so that the to-be-affected parameters return to more-normal values comprises: where the to-be-affected parameters are equal to one in number and referred to as a single to-be-affected parameter, for each pair of one or more distinct influencer/single to-be-affected parameter pairs in which the correlation value between the distinct influencer and the single to-be-affected parameter is high, adjusting the distinct influencer of the pair so that the single to-be-affected parameter returns to a more-normal value; where the to-be-affected parameters are equal to more than one in number and fluctuate similarly, selecting a first particular common influencer such that the correlation value between the first particular common influencer and each to-be-affected parameter is high; adjusting the first particular common influencer so that the to-be-affected parameters return to more-normal values; where the to-be-affected parameters are equal to more than one in number and fluctuate dissimilarly, selecting a second particular common influencer such that the correlation value between the second particular common influencer and each to-be-affected parameter is absolutely high, and such that the correlation value between the second particular common influencer and each to-be-affected parameter has a sign corresponding to whether the to-be-affected parameter is increasing or decreasing; adjusting the second particular common influencer so that the to-be-affected parameters return to more-normal values.

2. A method for automatically maintaining an autonomic computing system in a steady state, the autonomic computing system having a plurality of parameters, each parameter having one or more thresholds, the autonomic computing system having a plurality of influencers, adjustment of the influencers affecting values of the parameters, the method comprising: in response to determining that one or more of the parameters are each reaching one of the thresholds of the parameter, the one or more of the parameters referred to as to-be-affected parameters, identifying each to-be-affected parameter and the thresholds of the to-be-affected parameter; for each influencer, determining a correlation value between the influencer and each to-be-affected parameter; and, adjusting one or more of the influencers so that the to-be-affected parameters return to more-normal values, based on the correlation values determined, wherein the influencers that are each distinctly associated with the parameters are referred to as distinct influencers, such that adjustment of each distinct influencer affects the value of only one of the parameters at a response time, wherein the influencers that are each commonly associated with more than one of the parameters are referred to as common influencers, such that adjustment of each common influencer affects the values of more than one of the parameters at response times, wherein, for each influencer, determining the correlation value between the influencer and each-to-be affected parameter further comprises determining a correlation value between the influencer and the response time of the influencer as to each to-be-affected parameter, and wherein adjusting one or more of the influencers so that the to-be-affected parameters return to more-normal values comprises: where the to-be-affected parameters are equal to one in number and referred to as a single to-be-affected parameter, selecting a particular distinct influencer for which the response time is low, the correlation between the particular distinct influencer and the to-be-affected parameter is high, and the correlation value between the particular distinct influencer and the response time is low; adjusting the particular distinct influencer so that the to-be-affected parameter returns to a more-normal value; where the to-be-affected parameters are equal to more than one in number and fluctuate similarly, selecting a first particular common influencer such that a sum of the response times of the particular common influencer as to the to-be-affected parameters is low, the correlation value between the first particular common influencer and each to-be-affected parameter is high, and the correlation value between the first particular common influencer and the response time as to each to-be-affected parameter is low; adjusting the first particular common influencer so that the to-be-affected parameters return to more-normal values; where the to-be-affected parameters are equal to more than one in number and fluctuate dissimilarly, selecting a second particular common influencer such that a sum of the response times of the particular common influencer as to the to-be-affected parameters is low, an absolute value of a product of the correlation values between the second particular common influencer and the to-be-affected parameters is high, and an absolute value of a product of the correlation values between the second particular common influencer and the response times as to the to-be-affected parameters is low; adjusting the second particular common influencer so that the to-be-affected parameters return to more-normal values.

3. A method for automatically maintaining an autonomic computing system in a steady state, the autonomic computing system having a plurality of parameters, each parameter having one or more thresholds, the autonomic computing system having a plurality of influencers, adjustment of the influencers affecting values of the parameters, the method comprising: in response to determining that one or more of the parameters are each reaching one of the thresholds of the parameter, the one or more of the parameters referred to as to-be-affected parameters, identifying each to-be-affected parameter and the thresholds of the to-be-affected parameter; for each influencer, determining a correlation value between the influencer and each to-be-affected parameter; and, adjusting one or more of the influencers so that the to-be-affected parameters return to more-normal values, based on the correlation values determined, wherein the influencers that are distinctly associated with the parameters are referred to as direct influencers, such that adjustment of each influencer affects the value of only one of the parameters, wherein the influencers that are indirectly associated with the parameters are referred to as meta influencers, such that adjustment of each meta influencer directly affects one of the direct influencers, which in turn affects the value of one of the parameters, wherein, for each influencer, determining the correlation value between the influencer and each to-be-affected parameter comprises: for each direct influencer, determining the correlation value between the direct influencer and the parameter affected by the direct influencer; for each meta influencer, determining the correlation value between the meta influencer and the direct influencer affected by the meta influencer; and wherein adjusting one or more of the influencers so that the to-be-affected parameters return to more-normal values comprises: for each to-be-affected parameter, selecting a particular meta influencer for which a product of the correlation value between to-be-affected parameter and the direct influencer affecting the to-be-affected parameter and the correlation value between the particular meta influencer and the direct influencer affecting the to-be-affected parameter is high; adjusting the particular meta influencers so that the to-be-affected parameters return to more-normal values.